Re: [PATCH v2] memcg: Don't call schedule_work when no spinning is allowed

[Shakeel Butt <shakeel.butt@linux.dev>]

On Fri, Sep 04, 2026 at 03:15:54PM -0700, David Stevens wrote:
> On Fri, Sep 4, 2026 at 12:03 PM Shakeel Butt <shakeel.butt@linux.dev> wrote:
> >
> > On Fri, Sep 04, 2026 at 10:31:45AM -0700, David Stevens wrote:
> > > Memcg charging can be done from any context, but calling schedule_work()
> > > isn't safe from an NMI. If memory.high is breached from a context where
> > > spinning isn't allowed, use irq_work to schedule the reclaim work.
> > >
> > > Fixes: 3ac4638a734a ("memcg: make memcg_rstat_updated nmi safe")
> > > Acked-by: Michal Hocko <mhocko@suse.com>
> > > Signed-off-by: David Stevens <stevensd@google.com>
> > > ---
> > > v2:
> > >   - Added missing includes reported by Lorenzo and kernel test robot
> > >   - Added Acked-by
> > >
> > >  include/linux/memcontrol.h |  2 ++
> > >  mm/memcontrol.c            | 13 ++++++++++++-
> > >  2 files changed, 14 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> > > index 8170bb8066a2..4a5ef0aba475 100644
> > > --- a/include/linux/memcontrol.h
> > > +++ b/include/linux/memcontrol.h
> > > @@ -23,6 +23,7 @@
> > >  #include <linux/writeback.h>
> > >  #include <linux/page-flags.h>
> > >  #include <linux/shrinker.h>
> > > +#include <linux/irq_work_types.h>
> > >
> > >  struct mem_cgroup;
> > >  struct obj_cgroup;
> > > @@ -219,6 +220,7 @@ struct mem_cgroup {
> > >       spinlock_t       peaks_lock;
> > >
> > >       /* Range enforcement for interrupt charges */
> > > +     struct irq_work high_irq_work;
> > >       struct work_struct high_work;
> > >
> > >  #ifdef CONFIG_ZSWAP
> > > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > > index 6dc4888a90f3..0e8b302ca9ad 100644
> > > --- a/mm/memcontrol.c
> > > +++ b/mm/memcontrol.c
> > > @@ -62,6 +62,7 @@
> > >  #include <linux/seq_buf.h>
> > >  #include <linux/sched/isolation.h>
> > >  #include <linux/kmemleak.h>
> > > +#include <linux/irq_work.h>
> > >  #include "internal.h"
> > >  #include "swap_table.h"
> > >  #include <net/sock.h>
> > > @@ -2360,6 +2361,11 @@ static void high_work_func(struct work_struct *work)
> > >       reclaim_high(memcg, MEMCG_CHARGE_BATCH, GFP_KERNEL);
> > >  }
> > >
> > > +static void high_irq_work_func(struct irq_work *work)
> > > +{
> > > +     schedule_work(&container_of(work, struct mem_cgroup, high_irq_work)->high_work);
> > > +}
> > > +
> > >  /*
> > >   * Clamp the maximum sleep time per allocation batch to 2 seconds. This is
> > >   * enough to still cause a significant slowdown in most cases, while still
> > > @@ -2752,7 +2758,10 @@ static int try_charge_memcg(struct mem_cgroup *memcg, gfp_t gfp_mask,
> > >               /* Don't bother a random interrupted task */
> > >               if (!in_task()) {
> > >                       if (mem_high) {
> > > -                             schedule_work(&memcg->high_work);
> > > +                             if (allow_spinning)
> > > +                                     schedule_work(&memcg->high_work);
> > > +                             else
> > > +                                     irq_work_queue(&memcg->high_irq_work);
> > >                               break;
> > >                       }
> > >                       continue;
> > > @@ -4129,6 +4138,7 @@ static struct mem_cgroup *mem_cgroup_alloc(struct mem_cgroup *parent)
> > >               goto fail;
> > >
> > >       INIT_WORK(&memcg->high_work, high_work_func);
> > > +     init_irq_work(&memcg->high_irq_work, high_irq_work_func);
> > >       vmpressure_init(&memcg->vmpressure);
> > >       INIT_LIST_HEAD(&memcg->memory_peaks);
> > >       INIT_LIST_HEAD(&memcg->swap_peaks);
> > > @@ -4337,6 +4347,7 @@ static void mem_cgroup_css_free(struct cgroup_subsys_state *css)
> > >               static_branch_dec(&memcg_bpf_enabled_key);
> > >
> > >       vmpressure_cleanup(&memcg->vmpressure);
> > > +     irq_work_sync(&memcg->high_irq_work);
> >
> > On RT kernels, this will put rcu grace period here while we are holding the
> > cgroup_mutex. Easy fix would be to use IRQ_WORK_INIT_HARD instead of
> > init_irq_work() in mem_cgroup_alloc.
> >
> > Something like:
> >         memcg->high_irq_work = IRQ_WORK_INIT_HARD(high_irq_work_func);
> >
> 
> This executes as part of css_free_rwork_fn(), so it isn't under cgroup_mutex.

Oh yes, css_free_rwork_fn does not take cgroup_mutex. Though synchronize_rcu()
is still something to avoid but not a blocker.

You can add:

Acked-by: Shakeel Butt <shakeel.butt@linux.dev>